Skip to Content
Context Graph4. Explore the graph

Explore the graph

By the end of this page you will be able to read a codebase’s shape without opening a single file.

Prerequisites

  • An application with at least one completed ingest (Run an ingest).
  • Admin or Developer role.

The canvas opens on the architecture

Not on the file tree. Nodes are the things a system is made of — pages, API endpoints, services, models, background jobs, queues, databases — joined by typed edges: imports, calls, reads, writes, renders.

That is a deliberate choice. A file tree tells you how the code is filed; this tells you how it works.

The Context Graph canvas showing a multi-repository application's architecture

Narrowing what you see

A whole application at once is rarely what you want.

ControlWhat it does
RootCentre the view on one node. Leave it on all repos for the whole application.
DepthHow many hops out from the root to draw.
DirectionDepends on follows what the root needs. Depended on by follows what needs the root. Both does either.
FilterNarrow by name, type or repository.

Direction is the control worth learning. Depended on by answers “what would I break here?”, which is the question that usually matters before a change.

What the analysis found

Three cards sit beside the canvas, and they are findings rather than decoration.

Shape

Nodes, Edges, Repos and Unreferenced — a quick sense of size, and of how much is connected to nothing.

By type

The breakdown of node kinds. A system with forty services and two models usually means the models were not recognised, not that they do not exist.

Nothing refers to these

Code nothing else points at. Some is genuinely dead; some is an entry point the detectors did not recognise. Both are worth knowing.

Duplicated across repos

Type names defined in more than one repository. Often deliberate — a shared DTO copied rather than packaged — and occasionally the explanation for a bug where two copies drifted.

The Shape card beside the findings cards for unreferenced and duplicated code

Going from a node to the code

Open a node’s file in the Code tab to read the real source at the analysed commit.

ℹ️

The canvas shows one version. It is pinned to the commits that version was built from, so it does not reflect anything merged since. See Versions & diffs.

Common issues

  • “Run an ingest first” — No version exists yet.
  • The canvas is an unreadable tangle — Set a Root and a small Depth rather than viewing the whole application.
  • A node you expected is missing — Its language may have no symbol parser, or nothing references it and a filter is hiding it.
  • Everything looks unreferenced — Usually a single-repository application where the entry points are a framework the detectors do not recognise.